12

In order to ease supply issues, I am considering an overlapping footprint for the TQFP-32 and VQFN-32 packages of ATmega48/88/168/328 CPUs. The two packages have the same pinout. It's a commercial project, likely to make a few thousand units over a five-year period.

Has anyone tried this successfully? Are there any surprises? Any issue because of the signals going a little under the (larger) part? Any issues in fabrication or placement?

Obviously the alternative is to have to have multiple-version PCBs, with the additional documentation management overhead. This way obviously "wastes" the area under the chip for tracks.

It would look something like this (updated with thermals and GND/VCC):

enter image description here

Additional information in response to comments:

  • The design uses the internal oscillator
  • There are no analogue inputs or outputs
  • Most of the outputs are PWM; also RS485 at 250 or 500 kbit/sec; some i2c.
  • Bottom thermal pad will be connected to GND, with thermal vias, per datasheet p643. Pins 3, 5, 21 are also GND; pins 4 and 6 are VCC. The smaller part has a thermal pad on the bottom, the larger one does not.
Voltage Spike
  • 75,799
  • 36
  • 80
  • 208
jonathanjo
  • 12,049
  • 3
  • 27
  • 60
  • 2
    Analog input? Crystal oscillator? These might be adversely affected. Do be careful to terminate un-used pins. – glen_geek Nov 23 '22 at 13:39
  • @glen_geek thank you for your comment, edited question (internal osc, no analogue). – jonathanjo Nov 23 '22 at 13:44
  • 7
    Will you be using two paste masks? I'd be worried about the larger chip sitting on the solder paste deposited on the large QFN center pad and floating up when soldered. But I've never done this, so just hand waving here. – bobflux Nov 23 '22 at 14:07
  • @bobflux that's a *very* good point, makes me think two masks is a very good idea, most especially when larger package being installed. – jonathanjo Nov 23 '22 at 14:26
  • If the thermal pad is not connected to any PCB net, then it will float when the TQFP is installed. – RussellH Nov 23 '22 at 14:51
  • @jonathanjo Why would it be less difficult maintaining two different solder masks than two different PCB layouts? Do you mean two different paste stencils? Either way I'd ask your assembly contractor about their preference. My experience is that selective assembly/multiple BoM for one single PCB is almost always more trouble than it is worth. – Lundin Nov 23 '22 at 14:53
  • @Lundin, thank you, very helpful. I did mean two solder paste stencils, yes. Not two different solder masks on the PCBs: I agree, that's two different PCB designs. – jonathanjo Nov 23 '22 at 15:10
  • @jonathanjo Some contractors might have selective paste dispenser machines that can be used instead of stencils. Goes slower to apply that way, so it is ideal for smaller series. A few thousand units may or may not be considered a small series depending on the contractor. – Lundin Nov 23 '22 at 15:16
  • @RussellH thank you: edited to clarify that the centres is grounded. – jonathanjo Nov 23 '22 at 15:28
  • 2
    @Lundin I meant stencils too, I wrote some Frenglish by mistake – bobflux Nov 23 '22 at 15:42

1 Answers1

14

Has anyone tried this successfully?

My company does this in current designs because of Chip shortages and not just with this part. There are many companies that are doing this more and more. Or some of the ti buffer chips they can be stacked three times for three different sizes of parts.

One issue you might run into is if the outer part has a thermal pad under the part then that could potentially short pins out. You may have to tell the assembly house of your special condition but most should be able to handle it with just a loading diagram and a change in the BOM.

If there any signals over ~50 MHz you will be adding a very small amount of capacitance with the extra pads so this could affect rise times.

Voltage Spike
  • 75,799
  • 36
  • 80
  • 208